It's easy to see why the best modern day lenses have flatlined MTF charts, even at high spatial frequencies. The larger the spread of the image on a viewing surface, the more differences will be apparent, some maybe unexpected. Maybe a trend to more of a cine look, that iron fist in a velvet glove look.
rvh23 wrote:
I think you misunderstand. In Australia, it's very hard to return a lens to most vendors unless it is obviously faulty. I would never try to return a lens that I have adjusted in any way. For these GMs the only option I have when they are reasonable, but just not to my satisfaction, is to resell them usually at a loss despite having had essentailly no use.
In any case, I only tried adjusting one lens (and returned it to its original condition in the end) and I don't plan to do it with any other copies unless I know for sure I am keeping it and that it is minimally decentered. It's just too much effort.
...Show more →
It’s true that it’s hard to return a lens in Aus unless it’s obviously faulty, but this reluctance by vendors to let you do it is clearly inconsistent with Australian consumer law, which specifies not on,y that the product needs to be fit for purposes. It that it’s the consumer who decides that. The laws also specifies that it’s the consumer who chooses whether the manufacturer or the retailer deals with the return....so don’t be fobbed off by “you’ll have to approach Sony” or whatever by the seller.
I’ve found that the reluctance vanishes when you start to cite the law. But I agree it’s an off putting pain to have to do that.
Fred Miranda wrote:
Totally agree David, especially about your comment on less aliasing on A7R IV images. There is only one point I think we diverge... I think what's acceptable today may not be in the near future. Just speculating, perhaps very soon, it would be very cool to present our images in big 16K displays. Shooting with the R4 instead of R3 would give your detailed landscapes that extra 20% linear resolution -- unfortunatly mostly captured at center. That could be apparent to the most discerning audience but today, it's not visible even on big prints or 5K displays.
Either way, 42 to 61MP is not as significant as 24 to 61. That's an extra 60% linear resolution....Show more →
Yes I agree of course that if the 16k display is large enough, and if you get up close enough to it, you may be able to visually detect the difference in the centre of the image.
But I’m guessing you don’t think that kind of mild future proofing is the main reason to get a high resolution camera!
DavidBM wrote:
Yes I agree of course that if the 16k display is large enough, and if you get up close enough to it, you may be able to visually detect the difference in the centre of the image.
But I’m guessing you don’t think that kind of mild future proofing is the main reason to get a high resolution camera!
What's a high resolution camera? 10 years ago it was 20MP? 10 years from now it could be 200MP! My point is, get the highest resolution cam you can get for your detailed images in order to be as future proof as possible. Right now, the best bang for the buck is the A7R IV in my opinion. (Especially used)
Fred Miranda wrote:
What's a high resolution camera? 10 years ago it was 20MP? 10 years from now it could be 200MP! My point is, get the highest resolution cam you can get for your detailed images in order to be as future proof as possible. Right now, the best bang for the buck is the A7R IV in my opinion. (Especially used)
I agree that the A7R4 is the best bang for the buck at the moment, and well worth having.
But maybe what I mean by a high resolution camera is one whose resolution is beyond the level where it makes a visible resolution difference to the viewing of whole images.
Is that true of 60MP? In some circumstances, maybe no. But it will be at 200MP, though there will still be colour and aliasing benefits to 200MP if they are Bayer.
But I guess my original thought was more that the main benefits of higher resolution cameras in overall IQ are in areas where you probably aren't missing much or anything if your copy of a lens looks good at 1:1 on 42 MP, but you see some non-symmetry at 1:1 at 61 MP or above. Of course you might want to try and find a copy that is great viewed that way at 61MP in every corner, I've been knows to conduct that search, but it makes sense only if you sort of enjoy that. I think some people get stressed by it, thinking they must succeed in that search even if it's stressful, or else their images will look WORSE than if they had a lower resolution camera, or else that they will look no better in any way, and of course neither is true.
DavidBM wrote:
I agree that the A7R4 is the best bang for the buck at the moment, and well worth having.
But maybe what I mean by a high resolution camera is one whose resolution is beyond the level where it makes a visible resolution difference to the viewing of whole images.
Is that true of 60MP? In some circumstances, maybe no. But it will be at 200MP, though there will still be colour and aliasing benefits to 200MP if they are Bayer.
But I guess my original thought was more that the main benefits of higher resolution cameras in overall IQ are in areas where you probably aren't missing much or anything if your copy of a lens looks good at 1:1 on 42 MP, but you see some non-symmetry at 1:1 at 61 MP or above. Of course you might want to try and find a copy that is great viewed that way at 61MP in every corner, I've been knows to conduct that search, but it makes sense only if you sort of enjoy that. I think some people get stressed by it, thinking they must succeed in that search even if it's stressful, or else their images will look WORSE than if they had a lower resolution camera, or else that they will look no better in any way, and of course neither is true.
This is subjective though. For example, astrophotography may benefit from a symmetric and optimal lens. But, is it crucial for f/8-11 stopped down landscapes? Probably not...but everyone has their level of what's acceptable for them.
I have to confess that after testing 61MP vs 42MP, most of the improvement was at center and that will be true for higher megapixel sensors as well (like 200MP+). So it looks like we really need new sensor tech (like curved sensor) to start seeing off-axis advantages from shooting with even higher megapixel sensors. With the current sensor tech (or shape), it looks like the discrepancy between center and corner resolution will be even higher with higher megapixel sensors.
DavidBM wrote:
It’s true that it’s hard to return a lens in Aus unless it’s obviously faulty, but this reluctance by vendors to let you do it is clearly inconsistent with Australian consumer law, which specifies not on,y that the product needs to be fit for purposes. It that it’s the consumer who decides that. The laws also specifies that it’s the consumer who chooses whether the manufacturer or the retailer deals with the return....so don’t be fobbed off by “you’ll have to approach Sony” or whatever by the seller.
I’ve found that the reluctance vanishes when you start to cite the law. But I agree it’s an off putting pain to have to do that. ...Show more →
Wow, those are remarkable statements David. So if an Australian retailer tells you "we'll send it to Sony (or whoever) for evaluation" that is actually not required and you can insist the retailer accept the return regardless of what anyone else thinks of the lens? It sounds like the manufactuer's assessment in fact is irrelevant for any course of action you choose, and in theory at least, you could send your lens directly back to the manufacturer and ask for an exchange without evaluation. And Australia's crappy retailer return policies also are not actually legally binding? All sounds too good to be true...
But in practice I suspect you would very quickly wear out your welcome insisting on these rights here. You might be granted one exchange, two if you are very lucky, but after that would very likely find the lens is "out of stock". So if you need to try a substantial number to get a good sample, you may yet be out of luck.
Fred Miranda wrote:
This is subjective though. For example, astrophotography may benefit from a symmetric and optimal lens. But, is it crucial for f/8-11 stopped down landscapes? Probably not...but everyone has their level of what's acceptable for them.
I have to confess that after testing 61MP vs 42MP, most of the improvement was at center and that will be true for higher megapixel sensors as well (like 200MP+). So it looks like we really need new sensor tech (like curved sensor) to start seeing off-axis advantages from shooting with even higher megapixel sensors. With the current sensor tech (or shape), it looks like the discrepancy between center and corner resolution will be even higher with higher megapixel sensors....Show more →
61MP will also become diffraction limited before 42MP. There is a sweet spot where a very good lens will resolve better detail on the 61MP sensor, but it is a small window that decreases as the lens is stopped down.
pdmphoto wrote:
61MP will also become diffraction limited before 42MP. There is a sweet spot where a very good lens will resolve better detail on the 61MP sensor, but it is a small window that decreases as the lens is stopped down.
This is a common misconception. The effects of diffraction will be easier/quicker to see (due to the higher resolution), but diffraction limited...no. The amount of diffraction produced is the same. Now whether you see it or not is the question. F/stop for f/stop a higher resolution sensor will ALWAYS give you more detail than a lower resolution sensor of the same size.
In other words, a picture taken from an a7R IV at f/11 (insert lens) will always capture more detail than a picture from an a7R III at f/11 (insert same lens). It isn’t “diffraction limited”.
rvh23 wrote:
Wow, those are remarkable statements David. So if an Australian retailer tells you "we'll send it to Sony (or whoever) for evaluation" that is actually not required and you can insist the retailer accept the return regardless of what anyone else thinks of the lens? It sounds like the manufactuer's assessment in fact is irrelevant for any course of action you choose, and in theory at least, you could send your lens directly back to the manufacturer and ask for an exchange without evaluation. And Australia's crappy retailer return policies also are not actually legally binding? All sounds too good to be true...
But in practice I suspect you would very quickly wear out your welcome insisting on these rights here. You might be granted one exchange, two if you are very lucky, but after that would very likely find the lens is "out of stock". So if you need to try a substantial number to get a good sample, you may yet be out of luck....Show more →
Yeah you might wear out your welcome if you did it too often. Bits that’s true even of Amazon, I hear. What I do with big purchases these days is to use a retailer that I deal with often, and do some simple centring tests in store on a couple of copies.
greensony wrote:
This is a common misconception. The effects of diffraction will be easier/quicker to see (due to the higher resolution), but diffraction limited...no. The amount of diffraction produced is the same. Now whether you see it or not is the question. F/stop for f/stop a higher resolution sensor will ALWAYS give you more detail than a lower resolution sensor of the same size.
In other words, a picture taken from an a7R IV at f/11 (insert lens) will always capture more detail than a picture from an a7R III at f/11 (insert same lens). It isn’t “diffraction limited”.
I'm afraid that video you link to is an over-simplification of what can be a complex subject, and quoting it verbatim suggests that you don't fully understand sampling theory.
More sampling isn't always better or necessary. Taking the audio analogue, the Nyquist theory states that you only need two digital samples per analogue sine wave to reproduce the original exactly. That's why CD's are 44kHz sampling frequency, as that will faithfully reproduce frequencies up to 22kHz, which are already beyond the best persons hearing. All this nonsense about higher sampling rates more faithfully reproducing the original, is just marketing BS. The only exception to this is in the original master recording, where say 96kHz 24bit is worthwhile - to accommodate the losses in subsequent mixing of the track.
Fourier transform analysis tells us that images are really just a collection of different sampled frequencies. As with the audio analogue above, the sensor needs only to sample the highest spatial frequency that the lens can supply it with, up to the Nyquist frequency. Any further sampling will not provide anymore real information.
The bayer demosaicing complicates the normal Nyquist relationship sampling rate somewhat, and needs to be taken into consideration as well.
Diffraction is a physical property of light when it comes up against an edge. A perfect lens which has no optical aberrations will have a diffraction limited resolution at all apertures. However real lenses have optical aberrations, some of which are reduced by stopping the lens down: at the wider apertures, the diffraction resolution limits are so high relative to today's sensor resolution, that most of what you see degrading your image is these optical aberrations; as you stop down, some of these optical aberrations get reduced so the lenses image quality improves - the diffraction limited resolution is also dropping in the 'background' as well, but it still far exceeds the optical aberration limited resolution. Eventually, you reach an aperture where the diffraction limited resolution drops below the optical aberration resolution of the lens: from this point on the lens is considered 'diffraction limited'. This generally happens between f5.6-f11, depending on the quality of the lens.
Petegh wrote:
I'm afraid that video you link to is an over-simplification of what can be a complex subject, and quoting it verbatim suggests that you don't fully understand sampling theory.
More sampling isn't always better or necessary. Taking the audio analogue, the Nyquist theory states that you only need two digital samples per analogue sine wave to reproduce the original exactly. That's why CD's are 44kHz sampling frequency, as that will faithfully reproduce frequencies up to 22kHz, which are already beyond the best persons hearing. All this nonsense about higher sampling rates more faithfully reproducing the original, is just marketing BS. The only exception to this is in the original master recording, where say 96kHz 24bit is worthwhile - to accommodate the losses in subsequent mixing of the track.
Fourier transform analysis tells us that images are really just a collection of different sampled frequencies. As with the audio analogue above, the sensor needs only to sample the highest spatial frequency that the lens can supply it with, up to the Nyquist frequency. Any further sampling will not provide anymore real information.
The bayer demosaicing complicates the normal Nyquist relationship sampling rate somewhat, and needs to be taken into consideration as well.
Diffraction is a physical property of light when it comes up against an edge. A perfect lens which has no optical aberrations will have a diffraction limited resolution at all apertures. However real lenses have optical aberrations, some of which are reduced by stopping the lens down: at the wider apertures, the diffraction resolution limits are so high relative to today's sensor resolution, that most of what you see degrading your image is these optical aberrations; as you stop down, some of these optical aberrations get reduced so the lenses image quality improves - the diffraction limited resolution is also dropping in the 'background' as well, but it still far exceeds the optical aberration limited resolution. Eventually, you reach an aperture where the diffraction limited resolution drops below the optical aberration resolution of the lens: from this point on the lens is considered 'diffraction limited'. This generally happens between f5.6-f11, depending on the quality of the lens.
I’m afraid the theory you mention isn’t scientific fact of what can be a complex subject. I also didn’t quote anything verbatim, which suggests your attention to detail isn’t as good as you think it is.
You’re long winded last paragraph essentially reinforces my post and the video I linked, congratulations.
greensony wrote:
I’m afraid the theory you mention isn’t scientific fact of what can be a complex subject. I also didn’t quote anything verbatim, which suggests your attention to detail isn’t as good as you think it is.
You’re long winded last paragraph essentially reinforces my post and the video I linked, congratulations.
Good day, Mr. professor.
At 8:18 in the video he says and displays a text graphic "F/stop for F/stop a higher resolution sensor will ALWAYS give you more detail than a lower resolution sensor" - looks pretty verbatim to me.
You then go on to say in your own words "a picture taken from an A7rIV at f11 will always capture more detail than a picture from an A7rIII at f11. It isn't diffraction limited." This statement is false, as even the lower res A7RIII's sensor will be diffraction limited by the lens at f11.
Everything I said was based on the science as I understand it: if you know better, its customary to explain and provide evidence to the contrary, rather than just dismissing it, or turning personal.
If you genuinely wish to understand the topic better, the best lay-persons article I know of is the one from the Luminous Landscape's website here: https://luminous-landscape.com/do-sensors-out-resolve-lenses/
Fred Miranda wrote:
The way the Sony 12-24/2.8 GM rear filter slot works, glass filters won't fit since the filter must blend in order to go in. (only flexible gel/film filters work)
Very elegant solution from Haida!
Hi, are the Haida filters good quality ?
You seem to say that their thickness is too great? is it unusable?
What is the opinion of those who use them?
Thank you ;-)
MJBlackfoot wrote:
Hi, are the Haida filters good quality ?
You seem to say that their thickness is too great? is it unusable?
What is the opinion of those who use them?
Thank you ;-)
They have a good coating and the ND 3.0 has an exceptionally low color cast. But their thickness of 0.5mm induces additional FC and lowers slightly the corner sharpness even stopped as Fred has written. Stopped down the resulting IQ is still good enough for me (and I'm finicky regarding IQ). The handling is much better than any thinner alternative. Maybe the Aurora solution will solve this in the future. We'll see. I've tested them on the Sigma 14-24. You can find the test report here: https://wild-places.com/2020/01/06/rear-filters-for-sigma-2-8-14-24mm-dg-dn/
I received my fourth copy today, and it is the best one I've seen, with minimal decentering but some minor swing over the whole range that is masked by F4-5.6. After applying some temporary shims on the camera's mount it now looks fairly symmetric on my A7r4 even at 2.8, especially between 14 and 24mm. I may be able to improve 12mm symmetry with a bit more work, but even without that, when stopped down to 7.1 where I shoot focus-stacked landscapes it looks great. So I think I will call it quits with this copy.
In agreement with other reports, I clearly see best performance in the middle of the zoom range around 18mm, where it's phenomenally sharp all the way into the corners even wide open. Corner symmetry is also optimal at that focal length (nearly perfect), and worst at 12mm as also seen by several other FM members.
At 14mm I saw less corner asymmetry on average for the 4 Sigma 14-24 DG DN samples I tried last year than for these 4 GMs. But at 18mm it's been the other way around. At 24mm I didn't see a clear difference with these small numbers of lenses.
Although the Sony can be sharper than the Sigma at wide apertures if you have perfect copies of both, in practice it looks very much like the manufacturing tolerances are just not tight enough for that to translate to consistent real world outcomes. So if you plan to spend a lot more money to get the Sony rather than the Sigma, and expect that to translate to better IQ, you may be disappointed especially at 14-16mm where the Sigma is optimised.
rvh23 wrote:
I received my fourth copy today, and it is the best one I've seen, with minimal decentering but some minor swing over the whole range that is masked by F4-5.6. After applying some temporary shims on the camera's mount it now looks fairly symmetric on my A7r4 even at 2.8, especially between 14 and 24mm. I may be able to improve 12mm symmetry with a bit more work, but even without that, when stopped down to 7.1 where I shoot focus-stacked landscapes it looks great. So I think I will call it quits with this copy.
In agreement with other reports, I clearly see best performance in the middle of the zoom range around 18mm, where it's phenomenally sharp all the way into the corners even wide open. Corner symmetry is also optimal at that focal length (nearly perfect), and worst at 12mm as also seen by several other FM members.
At 14mm I saw less corner asymmetry on average for the 4 Sigma 14-24 DG DN samples I tried last year than for these 4 GMs. But at 18mm it's been the other way around. At 24mm I didn't see a clear difference with these small numbers of lenses.
10 years from now it could be 200MP! My point is, get the highest resolution cam you can get for your detailed images in order to be as future proof as possible. Right now, the best bang for the buck is the A7R IV in my opinion. (Especially used)